A bird’s eye on SARS-CoV-2 cases arround the world. An early world image as of April 2020.
Pin height represents magnitud of confirmed cases by country, .
Source: Own estimation based on offical WHO data compiled by Chennai, Tamil Nadu, India. Novel Corona Virus 2019.
The US data represents 32.2% of the total world’s cases, (April 25th, 2020) so it was removed from representation, to avoid overfitting.
The Latin American context
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
Brazil is the more affected country in the Latin American region by April 10th. Triplicating the number of cases that its predecessor, Perú reports. What kind of measures are these countries implementing? Are this countries appliying the same preventive controls? such as social distancing, border closings, mobility restrictions? The different control approaches may provide clues to better understand the virus spread rate.
México
Casos confirmados
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
Sonora
Casos confirmados. Early Stage. (May 20th 2020).
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data and Secretaria de Salud Sonora confirmed cases as of May 20th 2020.
While, absolute measures show important patterns, they only provide one side of the contagion process. We need relative indicators to control for population size in each area.
Are there any significant differences between patient´s age distribution between states?
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
The patient mean age is 41 years old, although no significant diferences across gender are found at this moment 2020-04-21, with an average age of 41 years for the Female population and 42 for Males.
We find that more Males than females are being infected. 58% vs 42% with 4 023 Females while 5 478 Males being reported since the first case was documented in the country, February the 28th, 2020.
Given the highly infectious nature of the virus and the labor market dynamics, a social determinant linked to a greater mobility for males, provides a research area that requires a formal empirical validation.
Age Distribution for covid19 deaths in Mexico.
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
The t test estimation identifies specific average patient age across states. Althought not all states exhibit a statistical diffence, we can observe an important fact, covid19 is mainly affecting productive age population in a range from 38 years old to 53. The spread of the CI’s indicates impacts in an important share of young population in some states, including Aguascalientes, San Luis Potosí, and Baja California Sur.
Source: Own estimation based on Dirección General de Epidemiología, official Mexican Government data.
Is the Covid19 growth curve flattening yet? Evidence from México.
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
Both lethality and growth rates from the 100th case threshold depic the evolution of the health emergency: The evidence for Mexico indicates that the emergency is taking a high toll at the country level. Although every state is been affected rather differently.
Taking a look at daily growth rates indicate a rather less critic scenario for the country, with signs of a decreasing trend, that already reached it’s peak at March 13th 2020. Update note as of February 10th 2022: However we now Know that there were several waves still coming. Those early figures were only the beginning of a long public health treath, that we are still figuring out two years later in 2022.
Daily growth rates for covid19 cases reported in Mexico.
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
The pattern suggests a slowing trend with an average daily growth rate of 16.1% considering the 46 days since the first case reported in the country. In addition, we can see an important lag in the public health authority’s reaction with a health emergency declaration made 17 days after the country’s covid19 cases reached its first peak daily growth rate.
What can be learned from Lethality rates at the state level?
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
Mexico´s age and sex adjusted Rates by State per 100 000 pop.
CDMX exhibits the highests rate, which occurs in Males. While the lowest is found in females from Chiapas.
Source: Own estimation based on Mexican Government official Open data.
It is found that CDMX exhibits the highest rate and is the male population that is been more affected.
Note: Estimates considers reported testing results as the target population for suspected COVID19 patients. The rate takes into account the proportion of positive cases to the total target population. Rates integrate the age and sex structure for standard population based on CONAPO 2020 official population estimates.
Outbreak curves: Mexico’s Northern Border States.
Source: Own estimation based on oficial Mexican Government data Dirección General de Epidemiología.
The covid19 confirmed cases path for the border states is showing a clear indication of different results of local containment measures for the spread of the virus. While, during the initial outbreak, Nuevo León (a state that host Monterrey, the third largest Metropolitan area in the country) had been the main focus of contagion at the border, the pattern dramatically changed by April 5th, when Baja California take off.
What kind of measures worked initially in Nuevo León, a state that apparently achieved an early flatten curve, what lessons can be learned from this state´s public health control efforts. And more importantly, what factors positioned Baja California as a rapid growth area. Update: As of February 202,2 we have evidence that the initial picture was only part of series of subsequent “waves” in which Nuevo León would continue to experiment a health crisis.
The 100th cases threshold.
Northern border states
Source: Own estimation based on Dirección General de Epidemiología, oficial Mexican Government data.
Baja California was the first border state to reach
the 100th case threshold, which occurred on April 06
2020.
Two days latter, the number of cases has doubled to 225.
Spatial distribution for confirmed cases at the Municipal scales.
Interactive Map: Activate the US-Mexico Border, Sonora, Arizona, or World layer below, to display data accordingly
Source: Own estimation based on datos abiertos.Dirección General de Epidemiología, oficial Mexican Government data. World and USA data compiled by Chennai, Tamil Nadu, India. Novel Corona Virus 2019 Dataset.Note: Daily Updated data.
Data source Warning: There are important differences in the number of cases reported by health authorities, particularly between the state level and the federal Goverment sources. The latter providing lower figures. See for example Sonora state, which is reportig 88 cases for 13/04/2020, while the oficial data base reports 72 cases for the same date.
US MX Border States: Relationship between confirmed cases and deaths reported.
Mexico
Source: Own estimation based on oficial Mexican Government Open data (OD).
US
Source: US data WHO offical data compiled by Chennai, Tamil Nadu, India. Novel Corona Virus 2019 Dataset.
Information Accuracy
The graph below determines differences in the number of cases reported by state in both official data bases. The accuracy gap is identified and classified as either under reporting or an over reporting state.
Source: Own estimation based on oficial Mexican Government Open data (OD) and daily technical communication CT.
During the referred period we found that on average 29.16% of the states in the country contained under classified data. This differences accounted to 349 covir19 cases not reported in the (OD) official source.
Furthermore, there is evidence of even larger differences when we consider the state reports vs. the Federal government official reports (OD). This is particularly evident at US-MX border Municipalities. Considering for Example the Sonoran border Municipality of San Luis rio Colorado, we found that the state is reporting 89 cases for April the 22nd, whereas the federal government via the official data mechanism designed (ODB), reports just 51 cases, an under report by 42%
From April 18th the OD, became the only official data government source…
Review case: Sonora.
National media start to report on the gap issue as early as April the 20th 2020. Six days latter the peak was reached and the information gap slowly moved downwards. Recently we observe not an increasing gap but rather a constant gap. To what extent this behavior reflects a “delay” in reporting? That’s an interesting question that was the focus of attention in the early days.
Modeling the outbreak path for confirmed cases México.
Note: This forecast model was developed for data available during the outbreak and reflects trends observed as data was being provided by official health authorities at the state scale in 2020. The model provides a historic portrait of the health emergency. The actual records for each concept may differ from updated versions of data management applications such as Covid-19 México provided by Mexican govenment institutions as methodologies for data collection and reporting practices are constantly updated.
Example of historic forecast given in April 2020.
Given the available information on the trajectory of the cases, we may expect to reach 32 286 cases by May the 10th 2020, with a lower bound of 27 791, and upper limit at: 39 109 as the worst-case scenario. It would all depend on, among other factors the measures implemented by health authorities in each state and also the growth rate on the US.
a posteriori note: The actual number of cases for this particular date was: 35 022. So considering the confidence interval it can be concluded that the model provided an accurate estimate of the trajectory of the health emergency.
An updated forecast using data available in 2022, indicated a prediction of 5 919 045 cases by April 8th 2022. However the actual number of cases was: 5 765 288. In this case for example the level prediction overestimated the actual scenario by 2.6%
So it can be concluded that time series models can be useful to provide accurate estimates and can be a reliable tool for planning and decision making when underlying assumptions are clearly indicated.
Disclaimer: This modeling exercise was developed to address an academic audience and to provide teaching materials suited for graduate students undertaking a masters in populations studies curriculum, so it should not be used as an official data source in any form**
References:
Data set compilation by Gabriel Alfonso Carranco-Sapiéns, based on official data from daily technical press reports. Github repository.